Skip to content

Retention & Cohort Methodology

How we build cohorts and measure repeat behavior across analytics.

Cohorts are monthly (loans) or quarterly (other modules) and assigned by earliest activity. For loans, an address joins a cohort in the month of its first funded loan in the requested role (lender or borrower). For sales and other modules, the cohort assignment is the quarter of the address's earliest indexed event of the relevant kind. Loans cohorts switched from quarterly to monthly because our post-spork backfill window is only ~120 days; a quarterly granularity would collapse all activity into one row.

Retention is computed as the % of a cohort that has at least one event of the relevant kind in each subsequent period. We do not require continuous activity — a member who returned in M3 but not M2 still counts as "active in M3." This is the "classic" retention curve and matches industry-standard SaaS cohort definitions. The cohort heatmap is labeled accordingly: "% active in month N," not strict retention.

Repeat percentages on the live KPI strip use a slightly different definition — a wallet is "returning" in the current window if it had any prior activity at any earlier date. This catches all re-engagement, not just consecutive-quarter retention.

Data sources

  • flowty_loans (Supabase)
  • sales (Supabase) — once Sales module ships

Refresh cadence

Every 10 minutes (loans) / 20 minutes (sales)

← Back to all methodology